Exchange Rate Movements and Foreign Direct Investments:

(1)

University of Groningen University of Uppsala Msc International Business Msc Economics & Business & Management

Exchange Rate Movements and Foreign

Direct Investments:

A non-linear exchange rate volatility approach and exchange rate competition

Gerber Norman Koster Master Thesis

March 2011

(2)

Abstract

This study focuses on the influence of exchange rate movements on United States (US) Foreign Direct Investment (FDI) outflows across 26 countries. It especially contributes to the literature by showing valuable insights in both the non-linearity of exchange rate volatility as in exchange rate competition for US FDI between neighboring countries. Firstly, by dividing the exchange rate volatility into eight subsamples it is found that exchange rate volatility has indeed a non-linear, U-shaped effect on US FDI flows. As a result, this study proves that countries adopting an extreme exchange rate regime (fixed or floating) are most likely to attract US FDI. Secondly, the current currency war illustrates the effect that relative exchange rate fluctuations probably have in the competition for US FDI. Investigation on relative competition revealed that several neighboring countries indeed compete for the same US FDI. Most important finding in this study is that exchange rate competition for US FDI is complex, not applicable to all countries, and to a large degree country specific.

(3)

List of Tables

Table 1. Empirical findings for the first hypothesis. ___________________________________ 323 Table 2. Lagging of variables. _____________________________________________________ 34 Table 3. FDI competition among neighboring countries. _________________________________ 37 Table A1. Different exchange rate regimes. ____________________________________________ 54 Table A2. Measurements of variables. ________________________________________________ 55 Table A3. Descriptive statistics. _____________________________________________________ 56 Table A4. Several comparable studies: Overview of variables. _____________________________ 57 Table A5. Singular neighboring countries. _____________________________________________ 58 Table A6. Included countries in this study. _____________________________________________ 58 Table A7. Correlation matrix. _______________________________________________________ 58 Table A8. Ramsey RESET. _________________________________________________________ 59 Table A9. Redundant Fixed effects test. _______________________________________________ 60 Table A10. Volatility band outcomes. __________________________________________________ 60

List of Figures

(5)

1. Introduction and problem statement

Competitive currency devaluation is currently receiving much attention as countries blame each other for distorting global demand and economic growth through intervening in monetary markets (The economist, 2010a). Monetary policy makers intervene by using quantitative easing methods (money printing, capital controls and currency intervention) to remain globally competitive. Brazil’s Finance Minister, Guido Mantega, even mentioned that a real “international currency war” has broken out (Webb, 2010). The basic idea behind competitive currency devaluation is that a relative lower exchange rate decreases the international price of a country its exported products, which enhances the growth of its battling economy to recover from the global downturn (Wheatley, 2010). Therefore it seems that exchange rates (both the level of exchange rates and its uncertainty) between countries are an important factor for a higher relative economic growth and more foreign capital. However, the currency war may undermine the global economic recovery when governments use exchange rates as a ‘policy weapon’ as Dominique Strauss-Kahn, head of the International Monetary Fund, warned (Kollewe, 2010).

The origin of this currency war lies in the opening of the financial systems (from the Bretton Woods system) in the 1970s. Whilst most countries maintained a fixed exchange rate regime in the years after the breakdown, the number of countries implementing a floating exchange rate system has been increasing over the last two decades (Kiyota and Urata, 2004). The adoption of a floating exchange rate regime can be seen as the next step in progressive liberalization of the financial systems. This liberalization caused a financial integration with global markets and became increasingly more important in the economic performance of many countries (Frenkel and Rapetti, 2010). For example, the economic performance of both developed and developing countries relies to a progressively larger degree on the amount of FDI received (Büthe and Milner, 2008). FDI inflows in developing countries’ economies counted for 1% of real GDP in 1970 and grew to 3.8% on average in 2007. For developed countries FDI inflows increased from 0.5% of real GDP in 1970 to 3.7% on average in 2007 (see Figure A1). Both increases indicate that FDI flows acquire a larger influence in the economic performance and growth of a country, which are now more influenced by exchange rate fluctuations.

(6)

could cover this uncertainty of exchange rate movements and its volatility on firm level would be hedging. However, several scholars suggest that international firms do not or cannot hedge against exchange rate uncertainty. Firstly, Carse, Williamson, and Wood (1980) discovered that firms that trade and invest internationally only cover their open positions between 15% and 30%, and therefore face currency risk. Although this research is aged, it may still be useful in highlighting that international firms are less capable of diversifying exchange rate risk than investors, that managing this risk does not add value to the firm, or that the foreign exchange rate effect is already factored in the firm’s market value (Sushko, 2007). Secondly, Bénassy-Quéré et al. (2001: 3) contributed by mentioning the difference between individual and corporate investors: “…while foreign individual investors should be indifferent to exchange-rate regimes as long as derivative markets allow them to hedge, foreign corporate investors should conversely worry about the exchange rate regime in the specific country, because they cannot hedge at their horizon. Moreover, corporate investors are interested in macroeconomic variables such as relative labor costs or purchasing power.” Lastly, McCarthy (2003) found that firms which are repeatedly exposed to exchange rate movements over a long period of time would not benefit from hedging activities. Therefore, the proposition is that international firms do not or cannot hedge against exchange rate movements.

Consequently, exchange rate fluctuations and its volatility indeed influence the FDI location choice and thus have an impact on the economy of a country. Frenkel and Rapetti (2010) likewise mentioned that the level of exchange rates have had a significant influence on the macroeconomic performances of many countries throughout the last few decades. This is due to an adoption of the floating exchange rate regime by those countries which automatically enlarged the fluctuations and volatility of its currency. In particular, an excessive appreciation of the exchange rate can have detrimental consequences for short and medium term economic growth in a country.1 As a result, it becomes more important for countries to be acquainted with the precise effect of exchange rate fluctuations on FDI inflows. Moreover, the revolution in information technology has enlarged the importance of exchange rates on FDI inflows as data is easily transferred and distributed across the world significantly faster. Organizations are therefore less dependent on a specific location for their investment and are more concerned about

1

(7)

the relative expected returns between countries (Chowdhury and Wheeler, 2008). These relative expected returns are directly influenced by exchange rate movements. These movements relate to both fluctuations in the level of exchange rates and the variability of exchange rates, known as exchange rate volatility. Hence, both the level and the volatility of exchange rates may influence the amount of investment (Chowdhury and Wheeler, 2008) and consequently the economic performance of a country.

During the last two decades many scholars have investigated the influence of exchange rate movements on FDI flows. In brief, studies have found that the correlation between the level of exchange rates on FDI inflows is generally negative, while the impact of the exchange rate volatility on FDI inflows is heterogeneous across countries, type of investment, and varies over time (Lin, Chen, and Rau, 2010). Hence, there is no consensus concerning the relationship of exchange rate volatility and FDI flows. However, Jeanneret (2010) discovered a possible non-linear relationship, which may explain the historic variety in the impact of exchange rate volatility.

This paper will investigate the influence of the level of exchange rates and its volatility on corporate FDI inflows from the United States (from now on US) and competition between countries across the world to receive FDI from the US.2 The US is chosen as the dependent FDI dispatching country for two reasons. First, the US is historically the largest single country that invests in countries across the world (CIA: World Factbook, 2009a). Second, while the area of FDI research is often obstructed by data constraints (Phillips, and Ahmadi-Esfahani, 2008), the Bureau of Economic Analysis (BEA) of the US Government reports FDI on a quarterly level (other databases report on an annual level). Quarterly observations allow for more accurate findings in the relationship between exchange rates and FDI flows. Figure 1 shows an overview of these quarterly US FDI flows to other countries from 1994 until 2010. The total amount of million US dollar FDI generally increases over time which is evidenced by the trend line. However, there is one major drop (even negative, during the 3rd and 4th quarter of 2005) which was probably due to the American Jobs Creation Act (October 2004) that encouraged repatriation of US firms overseas (IHS Global Insight, 2006).

2

(8)

Figure 1 . Total US FDI outflows.

The figure illustrates quarterly US FDI outflow observations during the period Q1 1994 – Q2 2010 in million US dollars. The dotted line represents a trend line (BEA, 2010).

The purpose of this paper is to investigate the relationship between the level of exchange rates and its volatility, and the received FDI for several countries across the world, based on US FDI outflows that are reported by the BEA.3 In this relationship the focus will be on two subjects: a possible non-linear relationship between FDI flows and exchange rate volatility, and the exchange rate competition for FDI. Therefore, the research question is stated as follows:

To what extent do exchange rate movements influence the amount of FDI received from the US?

This study proves that both the level of exchange rates and its volatility have a significant influence on the amount of US FDI received, although the influence is rather small compared to other variables. Moreover, the influence of exchange rate volatility is found to be non-linear and follows a U-shaped pattern. Exchange rate volatility has a negative influence on the amount of US FDI received for low volatility countries and a positive influence for countries with high exchange rate volatilities. Concerning exchange rate competition results indicate that competition for US FDI is country specific. Only some neighboring countries compete for the same potential US FDI with their relative level of exchange rates, while some others compete

3

(9)

with their relative exchange rate volatility. Still most investigated countries are not competing for US FDI.

With respect to these results this paper contributes to the literature in two ways. First of all, it has always been taken for granted that exchange rate volatility has a linear relationship with FDI flows. Jeanneret (2010) is the first who recently discovered a non-linear connection. By splitting exchange rate volatilities into several volatility levels the possibility to research a non-linear relationship is created, which may explain the non consensus in the academic literature (explained in more detail in Section 2.3). Therefore, this paper will investigate this possible non-linear relationship to reveal the real relationship between the exchange rate volatility and US FDI flows. Secondly, the competition between countries to attract foreign capital to encourage economic growth currently receives much attention. Since Bénassy-Quéré, Fontagné, and Lahrèche-Revil (2001) mentioned the importance of an exchange rate strategy in the competition for attracting FDI it seems that exchange rates are the main topic of debate regarding the global attractiveness of a country. Therefore, this paper investigates the influence of relative exchange rate fluctuations on US FDI inflows. Overall, this paper will contribute to a better understanding of the role of exchange rates on FDI inflows by providing several new insights, as both the non-linear relationship and the relative exchange rate movements are seldom studied yet in the academic literature. Moreover, comparing potential receiving countries is an essential factor when multinational enterprises (MNEs) consider relocation of facilities or outsource its production (Xing and Wang, 2006). For managers, investors, and institutions this paper will improve their process of making an international investment decision. In addition, governments and monetary policy makers of the investigated countries may get an improved insight of how their exchange rate policy affected, and will affect, FDI inflows from the US and possibly other countries.

(10)

2. Theory and hypothesis construction

This section will focus on the available literature of exchange rates, FDI flows, and the relationship between these two variables. Moreover, the hypotheses of this study are constructed during this section. Firstly, there will be an extended review of FDI, including the motivations of companies to invest in a certain country. Secondly, the available theory and the corresponding hypotheses on the level of exchange rates and its volatility will be explained in more detail in Section 2.2 and 2.3. Thirdly, Section 2.4 focuses on competition for FDI between countries and therefore reviews the literature on the influence of relative exchange rate fluctuations on the amount of FDI received in a country.

2.1 Foreign Direct Investments

The term FDI is defined as an international transfer of capital and can be interpreted in terms of comparison between expected returns on alternative investment decisions (Chowdhury and Wheeler, 2008). These alternative investment decisions are made by a continuum of heterogeneous companies (Jeanneret, 2010). Initially, a production company will produce its products domestically and sell them in competitive markets at home and abroad. Hence, the preferred first internationalization strategy for production companies is exporting (Gilroy and Lukas, 2006). The next step in this strategy is expanding production internationally, which includes an international transfer of capital. However, a service based company will directly transfer its capital internationally as production and consumption occurs at the same time.

As soon as a firm is pursuing internationalization of production or services (via FDI) there are several strategies following Perugini, Pompei, and Signorelli (2008): (i) market seeking; (ii) resource seeking; (iii) efficiency seeking; and (iv) strategic asset seeking.4 Companies expanding internationally by using a market seeking strategy only look for the exploitation of prior advantages from the host market, while some other companies try to exploit advantages based on low cost factors (resource seeking). The efficiency seeking strategy is characterized by companies seeking advantages from other externalities and host country characteristics. These three strategies are grouped as the ‘asset exploiting’ category by Castellani and Zanfei (2006: 86). The fourth strategy (and second category) ‘strategic asset seeking’ can be contrasted to the

4

(11)

previous three. In the first category, the foreign activities of companies are a substitute for their domestic activities, while companies in the second category search for complementary assets and are intended to combine those with their domestic assets. So, during the ‘strategic asset seeking’ strategy new assets may be created which can be beneficial for the host country firms as well.

As a result, motivations to invest in a certain country are not only based on the goals that a company sets, but are based together with characteristics of the host country itself and corresponding transaction costs. These three inter-related groups of motivations are described by the eclectic paradigm, better known as the OLI paradigm (Dunning, 2001). The first group is known as ownership-specific motivations (O) and includes foreign production activities that allow firms to achieve scale and scope economies (Perugini et al., 2008). The second group includes several motivations which rest on location advantages (L) such as reducing labor costs, exploiting technology/assets, agglomeration benefits and losing tariff barriers. Motivations of the last group are based on the internalization theoretical framework (I), which deals with the transaction cost-based theory of international production.

(12)

consensus as to which set of explanatory variables (determinants) should be used to understand FDI flows.

2.2 Level of exchange rates

Exchange rates move in relation to the exchange rate regime incorporated by monetary policy makers. Frenkel and Rapetti (2010) defined an exchange rate regime as the rules followed by the central bank of a country regarding to the degree of intervention in the foreign exchange market, and by the degree of official commitment in the determination of the nominal exchange rate. So the choice of a specific regime is dependent on the country’s individual circumstances and can differ over time (Frankel, 1999). In return, the implemented regime is dependent on the actual fluctuations of the exchange rate. The influence of these fluctuations on the amount of FDI received is mainly explained by the risk aversion theory and the real option theory.

However, the Bretton Woods system prevented exchange rate to fluctuate from the time of the Second World War until 1971. After the breakdown of this system, the exchange rates in an increasing number of countries have been fluctuating (Lin, Chen and Rau, 2006). Currently there are six singular exchange rate regimes distinguished by Frenkel and Rapetti (2010) that can be implemented by monetary policy makers (Table A1). Whilst intermediate regimes were widely used twenty years ago, Frankel (1999) and Calvo and Reinhart (2002) discovered a general trend towards the adoption of a fixed or (pure) floating extreme regimes.5 When adopting a floating exchange rate regime, the behavior of relevant macroeconomic variables, including inflation, balance of payments, output, employment, and the rate of economic growth are affected by exchange rate fluctuations (Frenkel and Rapetti, 2010). Regarding the balance of payments, many researchers have focused on the influence of exchange rate fluctuations on the amount of FDI received.

The first studies that linked FDI with exchange rates emerged in the late 1970s and at the beginning of the 1980s. Major contributions are made by Froot and Stein (1991) and Blonigen (1997) in the 1990s. Froot and Stein examined the connection between exchange rates and FDI flows by asserting that capital markets are exposed to informational imperfections. External financing is therefore more expensive than internal financing, so changes in wealth will lead to a change in demand for FDI. Froot and Stein (1991) concluded that host currency depreciations

5

(13)

will have a positive effect on inbound FDI. Blonigen (1997) focused instead on acquisition FDI and argued that exchange rate movements influence these FDI flows as acquisitions involve firm-specific assets which can generate different returns in different foreign markets. After a foreign acquisition the profitability of a multinational firm is exposed to exchange rate fluctuations. These fluctuations can affect relative asset valuations. Blonigen (1997) therefore expected an increase in FDI flows when the currency of the host country depreciates.

Several other studies have shown that a relationship exists between the level of exchange rate and the amount of FDI received or dispatched on a country level by using different theories. The two most important theories are: risk aversion and real option. First of all, the risk aversion

theory builds on the idea that exchange rate risk arises due to the time difference between the investment and actual profits (Phillips and Ahmadi-Esfahani, 2008), and that exchange rate fluctuations are uncertain. Bénassy-Quéré et al. (2001) contributed by noticing that risk-averse firms also search for alternative locations for FDI in order to reduce costs and minimize exchange rate correlations between locations. As a result, an exchange rate appreciation affects FDI flows negatively through both higher costs to exchange the investment into the local currency. Positive side effect of an exchange rate appreciation is the increase in local purchasing power of the population of foreign products. In particular, Bénassy-Quéré et al. (2001) stated that a FDI location decision is not independent from situations at alternative locations.

(14)

shock. In line with this is the idea that an appreciation of the host currency will lower the amount of FDI received in that country.

In the past decade, most other researchers concluded that a host currency appreciation will lower the amount of FDI received in the host country (e.g. Gorg and Wakelin, 2002; Feliciano and Lipsey, 2002; Kiyota and Urata, 2004; Bayoumi and Lipworth, 1998; Bénassy-Quéré et al., 2001; Osinubi and Amaghionyeodiwe, 2009).6 However, there are several studies that concluded there is no significant effect between the level of exchange rates and FDI flows (e.g. Matteson and Koo, 2002; Chakrabarti and Scholnick, 2000; DeVita and Abbott, 2007).

In summary, there are several regimes that form a base for exchange rate fluctuations. These fluctuations influence FDI flows and literature tends to show that an increase in the level of exchange rate (appreciation) reduces the amount of FDI received in the host country. Therefore, the first hypothesis is stated as follows:

Hypothesis 1: There is a negative relationship between an exchange rate appreciation in the host country and the amount of FDI received from the US.

2.3 Exchange rate volatility

The second issue of exchange rate movements on FDI flows is its volatility. Exchange rate volatility is defined as the variability of an exchange rate (Phillips and Ahmadi-Esfahani, 2008).7 This variability generally involves a great uncertainty in possible future outcomes, also known as risk as the variability is in nature unexpected. Exchange rate volatility affects FDI through three theories of which two are discussed previously: risk aversion of the organization and the option

value of investment flexibility (Lin, Chen and Rau, 2006). Foad (2005) described a third theory:

substitution FDI. First studies emerged in the late 1980s, but the impact of exchange rate volatility on FDI is still ambiguous (Kiyota and Urata, 2004), since results vary throughout the literature during the past decades. However, recently Jeanneret (2010) discovered a possible linear relationship between exchange rate volatility and FDI flows, which might explain the non-consensus in the literature.

6

Another interesting paper is Urata and Kawai (2000), who investigated the exchange rate relationship on an industry level and discovered mixed signs for different industries at Japanese enterprises.

7

(15)

First of all, the risk aversion theory explains that FDI decreases as exchange rate volatility increases (Osinubi and Amaghionyeodiwe, 2009). An increase in the volatility of an exchange rate will cause larger fluctuations in the expected return of an investment in that country, thereby creating further uncertainty over the actual future profits of a firm. Consequently, a risk averse firm will not invest in a country that has a relative high volatility, but instead will choose another country with a lower volatility to avoid this risk (Crowley and Lee, 2003). This results in a negative relationship between exchange rate volatility and FDI flows. Regarding to the volatility of exchange rates, companies intending to invest abroad will consider the characteristics between countries and not independently from each other.

Second of all, the option value approach can be used. Kogut and Kulatilaka (1994) mention that volatility of the exchange rate is one of the most important sources of uncertainty in an option value based approach.8 Considering the volatility of exchange rates the option value theory is slightly different, since an increase in exchange rate volatility might increase the investment value of waiting and not investing (Phillips and Ahmadi-Esfahani, 2008). Because of the option to defer the FDI activity (Lin, Chen and Rau, 2006) or change production to other facilities (production flexibility) it is possible to insure against disadvantageous exchange rate volatility developments, while maintaining the upside benefits of such an increase (Kogut and Kulatilaka, 1994). Result is an increase in the value of a multinational firm when exchange rate volatility increases. In general the assumption is that firms can adjust particular variables to new circumstances. Therefore, an option value approach argues that the value of an investing firm increases as volatility becomes larger, implying a positive relationship.

Last of all, Foad (2005) mentioned that the substituting FDI approach can explain a positive correlation between exchange rate volatility and FDI. Foad stated that FDI might be export substituting in times of volatile currencies. When volatility increases an international firm can decide to serve the host market via direct investment (in the form of a local production facility) instead of export the products, thereby protecting against exchange rate uncertainty when doing business. Hence, volatility is positive related to export oriented FDI. However, Foad (2005) mentions that a negative correlation is possible as well, as large exchange rate volatilities amplify the uncertainty of the actual profit. Moreover, firms that have several location choices

8

(16)

favor countries with a more stable currency. Consequently, countries with high exchange rate volatility will lose international capital (Foad, 2005).

So both theories and the approach of Foad have a different point of view and therefore different propositions about the impact of the exchange rate volatility on FDI flows. Hence, investigated correlations in academic literature are divided as well. One of the first studies that investigated the influence of volatility on FDI flows was Cushman (1988). Cushman concluded that high exchange rate volatility caused firms to invest in the host country to serve this market from within the host country. Hence, Cushman found a positive impact of exchange rate volatility on inward FDI for a host country. Secondly, Aizenman (1992) focused on the different exchange rate regimes. Results of this study demonstrated an investment flow to fixed regimes, as companies want to diminish their profit uncertainty, yielding a negative correlation. Thirdly, Goldberg and Koldstad (1995) concluded again with a positive correlation between FDI and exchange rate volatility in a model that considers profit maximilazation and export-substitution. Fourth of all, Kogut and Kulatilaka (1994) argued that a firm should choose to locate a plant in a country whose exchange rate is the most volatile to serve the population of the host country, as the substitution approach explains.

(17)

New is the non-linearity approach investigated by Jeanneret (2010).9 Jeanneret investigated the non-linearity relationship possibility of exchange rate volatility on FDI in OECD countries using a firm productivity perspective. Results showed that the exchange rate volatility coefficient is statistically significant, and is negative for low ranges of exchange rate volatility and positive for higher ranges. Jeanneret reasoned that low productivity firms react heavily negative on uncertainty when the level of volatility was low, since the increased volatility gives them the incentive to wait, as the payoff of the investment cannot increase in the future. On the other side, highly productive firms tend to export their products. The higher the volatility becomes, the more likely these firms are to relocate production abroad. Hence, highly productive firms react positive on an increase in uncertainty in terms of exchange rate volatility. By incorporating the motivations of firms with different productivity levels a non-linear, U-shaped relationship between exchange rate volatility and FDI is explained.10

In summary, several theories and approaches explain a different correlation of exchange rate volatility on FDI inflows. The diversity and non-consensus existing in the current literature might be the result of this distribution. However, it is possible that volatility has both a negative and positive effect on FDI in a U-shaped form. Hence, this paper investigates the possibility of a U-shaped relationship. Therefore the second hypothesis will be:

Hypothesis 2: There is a non-linear, U-shaped relationship between the exchange rate volatility in the host country and the amount of FDI received from the US.

2.4 Relative exchange rate fluctuations

A topic that received global attention during the past months is the attempt of countries to keep their currencies competitively low. Monetary policy makers intervene in the currency market in order to keep competitive to other countries in receiving international capital flows (The Economist, 2010a). The reason behind this is that monetary policy makers suppose that a lower-priced currency (i.e. a lower exchange rate) will boost global demand of its home products, increases foreign capital reserves, and hence, increase the economic growth of their country. Therefore, it can be worthwhile to intervene in the currency market. However, most countries

9

Two more papers explored the nonlinear relationship between exchange rate volatility and FDI before Jeanneret (2010): Wong (2005) and Eisenschmidt and Walde (2007). However, they focused on hedging benefits of options over forwards, whereas Jeanneret extended their work into the trade literature.

10

(18)

use the floating exchange rate regime (Calvo and Reinhart, 2002) and direct revaluation of its exchange rate is not possible. However, there are several methods to change the exchange rate indirectly by intervening in: (i) the interest rate, (ii) international asset portfolio, (iii) expectations, and (iv) order flows (Archer, 2005). First of all, by changing the real interest rate via especially unanticipated monetary policy actions the exchange rate tends to move. Secondly, a change in the relative scarcity of domestic versus foreign currency assets will cause a portfolio reallocation that changes relative prices in the process which can change the exchange rate. Thirdly, altering the future monetary policy expectations of market participants by using a signaling approach affects the exchange rate. Last of all, market professionals use order flow patterns to detect forces that are relevant to the exchange rate, and use this information to form exchange rate behaviors. These order flows can be altered by central banks with their own orders. In return market professionals react on this altered order flow, which changes the exchange rate indirectly.

Probably the best known example of currency intervention is currently China, where monetary policy makers have devaluated the Chinese currency substantially during the last two decades. It is even believed that the Chinese currency, the Yuan, is currently undervalued by about 40% (The economist, 2010b).11 Governments and policy makers of competitive countries believe that the current large amount of FDI that China receives is a result of this undervaluation of the Yuan. However, do relative exchange rates really have a significant influence on FDI flows across countries?

The first study that mentions the importance of controlling cross-country exchange rates in order to receive capital inflows is Bénassy-Quéré et al. (2001). In their opinion foreign companies consider the relative price and costs between countries before they decide to invest at a certain location. The idea is that foreign companies prefer to invest in the neighboring country when its currency is relatively cheaper due to a lower exchange rate. Therefore, these companies incorporate relative exchange rate fluctuations into the decision process where to invest. Bénassy-Quéré et al. (2001) included this proposition into their model by highlighting the interaction between host countries through their exchange rate regime. They found for positive correlated currencies that a relative depreciation of country j (decreased competitiveness)

11

(19)

reduces FDI to country i via a substitution effect. On the other side, in the case of negative correlated currencies FDI to country i will raise during a relative depreciation of country j via a diversification effect. So a relative fluctuation in exchange rates alters the competitiveness and consequently the distribution of FDI flows between countries.

Secondly, Barrell, Gottschalk, and Hall (2004) cited that FDI flows are mainly determined by the relative factor cost competitiveness of the host country. A foreign direct investor decides its location choice by considering the potential host relative competitiveness, which is proxied by the relative real exchange rate of the potential host country against the real exchange rate of the investor (Barrell, Gottschalk and Hall, 2004).

Another study that focused completely on the impact of exchange rates on competitiveness across countries was done by Xing and Wang (2006). They investigated the role of exchange rates and to what extent they might have altered the relative competitiveness of the recipient country for FDI from the same source country within Asia, since geographic distribution of Japanese FDI has changed substantially across Asian countries, in favor of China. Relative exchange rates among Asian countries were likely to be the driving force behind the altered FDI distribution, since the Chinese Yuan devaluated in relation to ASEAN-4 countries over a long period.12 The results showed that relative FDI is determined by the relative real exchange rate between the host and source country. Therefore, Xing and Wang (2006) extended the knowledge of how exchange rates influences FDI flows by showing that exchange rates play a significant role in shaping competition among potential FDI recipient countries.

The current attention of global currency competition across countries, together with the previous discussed studies, assumes that a relative appreciation of the exchange rate of one country will lower its competitiveness for FDI and therefore leads to a relative decrease in FDI in relation to other countries. The third hypothesis is based on this construction and is described below:

Hypothesis 3: A relative appreciation of the currency of country i in relation with country j will cause the foreign direct investment received from the US in country i to decrease in comparison to country j.

12

(20)

3. Methodology

This section will clarify the methodology used to examine the hypotheses of this study. The first section will introduce the general methods used. Next, Section 3.2 deals with a complication of the dependent variable and Section 3.3 will clarify the methodology used for the first two hypotheses. The following section describes all included diagnostic tests and there is a separated area (Section 3.5) that accounts for the control variables included in this study. Last of all, the methodology for hypothesis three is explained in Section 3.6.

3.1 Methods used

When investigating the determinants of FDI flows many researchers use econometric models. Most applied model is a single linear ordinary least square (OLS) regression of aggregated FDI flows and some explanatory variables across time series or with the use of panel data (Phillips and Ahmadi-Esfahani, 2008). Of these two alternatives the panel data, combined with the standard ‘gravity’ framework is the primary empirical model used with international trade and investments (Lizardo, 2009). This gravity framework basically explains trade or FDI flows based on several economic and distance variables between two units (Blonigen, 2005). Moreover, the gravity model has several benefits over other models. First of all, Phillips and Ahmadi-Esfahani (2008) state that the gravity model is associated with a certain standard set of determinants, allowing for both distance and country effects.13 Secondly, the gravity model is recognized for its consistent empirical success in explaining FDI flows (Bergstrand, 1985) together with exchange rates. Last of all, as Clark, Tamirisa, and Wei (2004: 63) cited: ‘The gravity model has proved to be robust and successful in a wide variety of empirical applications. Moreover, the gravity model has a strong foundation in international trade theories, from those based on country differences in factor endowments or technology to models of increasing returns to scale and monopolistic competition.’

Panel data generally consists of cross-sectional data observed though multiple time periods. Within panel data fixed or random effects might be present, based on the nature of the data, which can alter the outcomes of the study. The redundancy test is formally used to confirm the existence of these effects in the dataset. Next, when significance is found the Hausman test will explain whether these effects are random or fixed. Random effects are normally assumed when

13

(21)

individual countries in the dataset are randomly chosen to represent a larger group of countries. In contrast, fixed effects assume that all behavioral differences over time and between countries are captured by the intercept itself (Hill, Griffiths, and Judge, 2001: 20). A significant result of the Hausman test would mean that the panel data model with fixed effects is the appropriate model.

3.2 FDI flow complication

Observations of the dependent variable, US FDI flows, vary widely across time and between countries and should be standardized to isolate statistical errors in the measured data Additionally, an OLS regression model assumes normal distribution of the residuals of the dependent variable. Normally, basic logarithms are used to standardize the dependent variable. However, reported US FDI flows of the BEA include negative observations. Negative FDI flows are possible, since FDI data flows are often presented on a net basis. Hence, a negative sign indicates that at least one of the three components of FDI (equity capital, reinvested earnings or intra-company loans) is negative and not offset by positive amounts of the remaining components (UNCTAD, 2002). These are examples of reverse investment or disinvestment.

As a result, some researchers simply exclude these negative observations (Levy-Yeyati, Panizza, and Stein, 2003), although negative observation may contain important information (i.e. negative observations are more likely during political instability, or economic downturn). Therefore, excluding those observations could bias the results. However, negative (FDI) observations cannot be standardized with basic logarithms. A solution to retain these negative FDI flows is a combination of Levy-Yeyati et al. (2003) and Osborne (2002) proposals to transform data for normalization of variables. Formula 1 shows how the dependent FDI variable is consequently transformed to fit for this research:

(

)

( )

[

]

*100

)

(FDI Log FDI c Log c

Log _it = _it + − (1)

i = 1, …, 26 ; t = 1, …, 44

(22)

the negative values back to its original sign the logarithm of the constant will be deducted again. Finally, the transformed FDI data will be multiplied by hundred to bring those values back to a higher range in line with about the same proportions as basic logarithms.14 It is important to mention that this transformation does not change the relative order of the data (Osborne, 2002). Additionally, the original actual FDI flows and transformed FDI flows have a correlation of 0.99922 with each other as shown in Figure A2.15 However, because of the transformation of the dependent variable, differences in interpretation should be considered when interpreting the empirical results (Brooks, 2008: 176). With the logarithm, data is now curvilinear and therefore compresses one part of the observations more than another (Osborne, 2002).

3.3 Linear regression model

For this study the real exchange rate will be used to investigate the influence on US FDI flows. The real exchange rate is defined by the purchasing power parity as the nominal exchange rate, adjusted by the ratio of foreign to domestic price level (Kipici and Kesriyeli, 1997). When the purchasing power parity does not hold the real exchange rate will take differentials in the price levels among countries into account. Due to this adjustment is the real exchange rate superior to the nominal exchange rate. Moreover, the real exchange rate can be used in the field of competition for foreign trade and FDI across countries (Kipici and Kesriyeli, 1997). The following two formulas explain the construction of the exchange rate variables (Lizardo, 2009):

(

)

[

i US i

]

_t it e P P ER = / (2)

(

it

)

it VAR ER ERV = (3)

For formula 2, the real exchange rate,

[

ei

(

PUS /Pi

)

]

_t, represents the ratio of the price in the US

(P_US), to the price in country i (P_i), and (from an American perspective) the indirect nominal exchange rate quote, ei, during period t.

16

The left hand side variable,ERit, is defined as the

amount of foreign currency of country i required to purchase one US dollar during the same period t. The volatility of the exchange rate in country i during period t,ERV_it, is specified in

14

By multiplying with 100 the relative distribution will not change.

15

Outliers in the dataset are already taken out in this correlation graph (see Section 3.4 for outlier specification)

16

(23)

formula 3. Here the volatility of the real exchange rate is stated as the variance of the real exchange rate. As floating exchange rates are close to random walks, the volatility variable will include both the predictable and unpredictable parts of the exchange rate fluctuations (Goldberg and Kolstad, 1995).

Secondly, the influence of the basic relationship of the exchange rate level and its volatility to the amount of US FDI received will be examined. As can be derived from Section 3.1, an ordinary least square (OLS) single regression panel data approach, together with the gravity model and fixed or random effects will be used. The regression formula used is quite similar to that of Bénassy-Quéré et al. (2001), Crowley and Lee (2003), Chowdhury and Wheeler (2008), and Lizardo (2009). The dependent FDI variable, independent exchange rate variables and several control variables are formulated in the following formula:17

(

)

(

)

(

)

∑

(

)

= + Γ + + + = k it it it it

it Log ER Log ERV

FDI Log 1 $ # 2 1 0 α α α ε α (4) i = 1, …, 26 ; t = 1, …, 44 ; # = 3, …, 11

The left-hand-side variable is the logarithm of the amount of FDI from the US to country i in period t. For the right-hand-side the logarithms of the independent variables average level of real exchange rates for each period, ERit, and exchange rate volatility, ERVit, are included. The last

part of the single regression formula shows a sum of all the included control variables, Γ , plus _it an error component,ε_it. The control variables used in this research are explained in Section 4.5. Additionally, the free parameters,α , show the constant and coefficients of each variable.

Next, the non-linearity of the exchange rate volatility will be investigated (hypothesis 2). As Jeanneret (2010) explains for testing this relationship, the data has to be separated into several volatility bands of which each band corresponds to a certain range of exchange rate volatility. Chosen is for eight bands to compare the results of Jeanneret (2010), where eight volatility bands were used as well. Each subsample is investigated separately for the influence of exchange rate volatility on US FDI, by using regression formula 4. After that the eight resulting coefficients will be combined and plotted into Figure 2 (page 36) for a graphical illustration. A specific explanation regarding each volatility band is added in Table A2.

17

(24)

Lastly, the option value theory includes the possibility to account for delay effects of exchange rates on FDI flows. Since firms are able to invest, wait, or not invest it is possible that a firm decides to invest after they recognize an exchange rate shock has occurred. Moreover, there is a time difference between the exchange rate movement, the decision to invest and the actual investment. Lagging of time periods might therefore help to explain whether it takes a certain period before a shift in the exchange rate and its volatility has an influence on FDI flows.18 This lagging approach is implemented in formula 4 (see formula 5). Both independent variables as all control variables will be lagged. Additionally, the appropriate rounded lag length for this study is four periods (see Table 1) based on the Hannan-Quinn criterion.19 One additional period is added to control for the influence of all variables on FDI flows at a longer lagging period.

(

FDIit

)

=α +α

(

ERi(t−η)

)

+α

(

ERVi(t−η)

)

+

Log ₀ ₁log ₂log

( )

(

)

∑

= − + Γ k it t i 1 $ # ε α η (5) i = 1, …, 26 ; t = 1, …, 44 ; # = 3, …, 11 ; η=1, …, 5 3.4 Diagnostic tests

Several diagnostic tests will be used to control or improve the fitness of the dataset for investigation. First of all, all variables in the dataset are controlled for the degree of linear association, better known as multicollinearity, by using a correlation matrix (Brooks, 2008: 171). In general, when a correlation between two variables exceeds 0.70 it is very likely that both variables measure the same variation.20 Therefore, I will define a 0.70 correlation as maximum degree of linear association between variables. When a higher correlation is measured, one variable will be excluded during the investigations, based on how often those variables are used in comparable studies.

Secondly, the OLS regression model relies on the assumption that errors in the regression are normally distributed (Brooks, 2008: 163). The transformation of the dependent variable

18

I actually investigated the delay effect within the non-linearity of the exchange rate volatility as well, but due to the small subsamples and consequences of delaying those samples no proper investigation was possible.

19

The Hannan-Quinn criterion was found to outdo other criteria in a study of Liew (2004). Additionally, the Akaike Info Criterion and Schwarz Criterion found an optimal lagging period of 3.43 and 3.69 periods respectively.

20

(25)

standardized the data and whether residuals of the dependent variable are now normally distributed is tested with the Jarque-Bera test. The Jarque-Bera test confirms whether residuals are normally distributed by incorporating the skewness and kurtosis of the sample.21

Next, non-linearity in the regression is tested by using the Ramsey Regression Equation Specification Error Test (Ramsey RESET) (Brooks, 2008: 177). The test formally checks whether (possible) non-linearity of explanatory variables have significant influences in explaining the estimated value of the dependent variable. Insignificance of the general Ramsey RESET test implies a linear relationship with the dependent variable.

Lastly, outliers of the dataset are determined to improve the explanatory power of the regression results. Observations are in considered as outlier when the observation does not fit with the remainder of the data (Brooks, 2008: 164). Outliers are measured following the ‘three-sigma rule’ principle (Ruan, Chen, Kerre, and Wets, 2005: 318), which explains that an observation is considered as an outlier when its value is more than three times the standard deviation from the norm (the mean). Table A3 includes the maximum and minimum values based on the ‘three-sigma rule’ principle.22 A dummy variable is used to mark all outliers to expel them from main investigations (Brooks, 2008: 165), but are still investigated in a separate model to check for relevant influences. All these outliers in the dataset are defined as 0, other observations as 1. This dummy variable, dFDI_it, will be implemented as like any other variable in the regression model (Brooks, 2008: 169) of formula 4 (see formula 6). Consequence is that the dataset is unbalanced without these outliers.

(

)

∑

(

)

= + Γ + + + = k it it it it dFDI FDI Log 1 $ # 3 0 ... α α ε α (6) i = 1, …, 26 ; t = 1, …, 44 ; # = 3, …, 11 3.5 Control variables

Several extra variables are included in the regressions to control for the effect of the dependent variable. Included control variables in this study are based on their usefulness of previous studies and the gravity model that is used. An overview of several studies of last decade and the control variables included in those studies are attached in the appendix (Table A4). Next,

21

Skewness stands for how symmetric the residuals are around its mean. Kurtosis is a measure for the peakedness of the normal distribution.

22

(26)

the variables that are included in this research are formulated in formula 7.23 In Table A2 (appendix) there is also a brief overview of all the variables and how they are measured.

(

)

+ +

[

(

)

]

+ = Γ

∑

= it it it k

it) log GDP CPI log DIST GEO

( 3 4 5 1 $ # α α α α α6DIST

(

KN

)

it +α7DIST

(

LF

)

it + α₈OPENNESS_it +α₉TAX_it+α₁₀IR_it +α₁₁ED_t (7)

MARET SIZE: In the gravity model the variable market size is the most widely used control variable and often proxied by their respective real GDPs (Sekkat and Galgau, 2002). This GDP variable, GDP_it, is constructed as logarithm of the total GDP of the host country in millions of current US Dollars. In the literature a larger market is likely to receive more FDI. Therefore, a positive relationship is expected.

RELATIVE CPI: Lin et al. (2006) notice that it has been suggested that expensiveness of a production location is one of the important driving forces of FDI. A measurement that accounts the expensiveness of production in the host country could be a comparison between national wages. However, national wages are measured differently in each country: per unit, per hour, per week or even per month. Some countries even have only wage data of the manufacturing sector or no data at all available. Therefore a different approach is needed, in order to include a consistent proxy that accounts for the expensiveness of a country. Chowdhury and Wheeler (2008) use the general price level in each country as that proxy. The Consumer Price Index (CPI) is a standardized proxy that measures the general expensiveness of a country, and will therefore be used during this study. For this study, the CPIs of all countries are transformed relatively to the US 2005 market prices and are included as variable CPI_it. The relationship is expected to be negative, as a drop in the index would mean lower relative costs for the US subsidiary, which will in return increase its production in this country (Foad, 2005). Also Chowdhury and Wheeler (2008) found that US FDI flows are negatively related with the CPI.

DISTANCE (GEO, KN, and LF): In this study there will be three distance control variables, of which two are proxies. First of all, the geographical distance variable,DIST

(

GEO

)

_it, is accounting for the actual distance between the US and the FDI receiving country. As Sekkat and Galgau (2002) note there is in a negative relationship explained in the literature, based on the

23

(27)

fact that location production at a greater distance implies more transportation costs. Secondly, two distance proximity measures are included from the paper of Nachum, Zachum, and Gross (2008): proximity to the world knowledge centers, DIST

(

KN

)

it, as measured by the proximity to

the world its research and development (R&D) investments, and proximity to the world market centers,DIST

(

LF

)

_it, as measured by the proximity to the world its labor force.24 The importance to take these variables into account, and especially the first one, was shown in the regression results of Nachum et al. (2008). Here, based on the gravity model that FDI decreases when distance increases, a negative relationship is expected for both variables.

OPENNESS: This control variable indicates the level of integration of the recipient country with the global economy. A high openness level would explain a higher amount of FDI inflows, since a trade promoting country tends to trigger export-oriented FDI (Xing and Wang, 2006). Openness is measured as ratio of the sum of total exports and imports of country i, to the GDP of country i (Bénassy-Quéré et al., 2001). All the measures are in current dollars for each period. Logically, the expected sign of OPENNESS_it is positive, as completely closed economies have no FDI inflows.

TAX: Both international as public economists believe that higher taxes discourage FDI flows (Blonigen, 2005). Blonigen mentions in his paper that taxes have a direct impact on the profits of foreign affiliates, since foreign affiliates cannot avoid the foreign taxes on retained earning.25 Therefore, the height of a host country’s tax rate will influence the decision of a US corporation to invest in this country. For all countries the tax rate, TAXit, is measured by the corporate tax

rate applied per period in that country. The expected sign is negative, as it is likely that US companies invest less in a country with higher corporate taxes and those taxes directly lower US host retained earnings.

INTEREST RATE: This control variable is based on the cost of capital for corporations and foreign affiliates in the host country. While international corporations might lend their capital on the international capital market, it is possible that the subsidiary of the international mother corporation has to lend its capital from a national bank. Therefore, the host country interest rate

24

Higher values imply a lower proximity. This is important to know when interpreting the results, because a negative coefficient indicates a positive effect of proximity.

25

(28)

might influence the decision to invest in a specific country. Significance of this control variable would imply that foreign affiliates indeed lend capital on the national capital market, and use this rate in their FDI location decision. The interest rate, IRit, stands for the average real interest rate

charged by commercial banks in the host country. Just like the host tax rate a negative sign is expected here, as higher local interest rates mean higher costs for the subsidiary (when borrowing on the national capital market).

ECONOMIC DOWNTURN: The last control variable considers the influence global economic downturns might have on US FDI flows, by using a dummy variable. Following the IMF a global economic downturn, ED_t, is at hand when the global economic growth falls below 3 percent. During the period 1999-2010 there are three individual recessions with this condition (see Figure A3). Therefore, the periods Q1 2001-Q2 2002, Q1 2003-Q2 2003, and Q2 2008-Q4 2009 have a value of 1, which implies an economic downturn period. All other time periods have a value of 0. An economic downturn is likely to have a negative influence on the amount of US FDI flows, as US companies have less capital to invest abroad.

3.6 Relative exchange rate fluctuations

(29)

regressions with formula 8.26 Table A5 includes all the thirty country combinations that share a border with each other.27

The methodology for this hypothesis is basically the same as the linear OLS regression model (formula 4), with difference that relative observations between countries is used. Xing and Wang (2006) applied formula 8 to determine the influence (δ₁) of the relative exchange rate fluctuation on the relative difference in the amount of FDI received. However, some control variables cannot be included (like geographic distance and tax rate), as these are mostly static observations and do not contain enough variation to be implemented into this regression.

+         +         +         + =         t j i t j i t j i t j i GDP GDP ERV ERV ER ER FDI FDI 3 2 1 0 δ δ δ δ it t j i t j i t j i IR IR Openness Openness CPI CPI ε δ δ δ _ +       +         +         6 5 4 (8)

Where FDI denotes the amount of US direct investment in million US dollars and the independent variables ER and ERV are respectively the average real exchange rate and exchange rate volatility. The remaining variables control for the influence of GDP, which is the gross domestic product denoted in million US dollars, CPI is the relative consumer price index,

Openness measures the level of economic openness in the recipient country, and IR stands for the interest rate.εit is the error component. Subscript i stand for a country that is compared with its

neighboring country j during period t. For each investigation another combination of two countries is examined for the relative influence on US FDI.

4. Data description

As explained before, this study focuses on US FDI outflows to the countries that are reported by the Bureau of Economic Affairs (BEA).28 BEA reports US FDI outflows every quarter of a year to 58 receiving countries since 1994. The research is based on the influence of different exchange rates on the amount of US FDI received. A consequence of this criterion is that only

26

Due to time limitation of this study not all possible combinations of countries could be compared and investigated for (global) competitiveness. Future research might focus to the other possible combinations.

27

Border connection can be on both land and water. Additionally, the European Monetary Union was specified as one region (with Germany as representing country). Germany has therefore a shared border with Switzerland, the United Kingdom, and Norway during this research.

28

(30)

Germany, out of all European Monetary Union members, will be included. Germany is chosen as European Union country since Germany is the largest economy in Europe and a leading FDI recipient (CIA: World Factbook, 2009b). Furthermore, due to scarcity of data, which is the greatest constraint that researchers face in this topic (Phillips and Ahmadi-Esfahani, 2008), only 26 countries have sufficient data to be included for further investigation. The included countries are stated in Table A6.

For the observation period is chosen for the timescale between Q1 1999 and Q4 2009 (a total of 44 observation periods), as the introduction of the Euro occurred 1 January 1999.29 By starting at this date the implementation effects of the Euro is not affecting the results during this research. However, global economic downturns, like the financial crisis, might still influence exchange rates and as a consequence the results. The dummy variable,EDt, is added to control the

influence of these economic downturns.

The independent variables, the real exchange rate level and its volatility, are completely retrieved from the International Monetary Fund: International Financial Statistics (IMF:IFS) database (www.imfstatistics.org), by using the Thomson- and WM-Reuters. The raw data was retrieved daily, and converted into quarterly figures to fit with the other variables.

Quarterly data for all control variables come from several sources and databases. First of all, data of GDP, trade and population size come from IMF:IFS (GDP data of Egypt, Singapore, and Taiwan are retrieved from the national databases),30 IMF:IFS, and DataStream respectively. Secondly, the geographical distance variable is calculated with Google Earth and the paper of Nachum et al. (2008) is used for the control variables distance to knowledge and labor force. The tax rates are mostly retrieved by the World databank: WDI and GDF database (www.worldbank.org). However, because of some missing data also databases of Deloitte (Deloitte, 2010), KPMG (KPMG, 2009), Tax Policy Center (Tax Policy Center, 2010), and the OECD (stat.oecd.org) are used to retrieve tax rate information. Furthermore, information of the Consumer Price Index is retrieved from the IMF:IFS database and converted relatively to the US CPI and global prices by using the International Comparison Program results (2008) of the

29

Smaller time periods would be even better, since exchange rates fluctuate on a daily base. However, FDI flows occur less frequently and are better observable on a monthly level. Therefore, monthly data would be the best observation level. However, most variables (including FDI flows) are only reported annually or quarterly.

30

(31)

World Bank. Last of all, information about the actual interest rate is also retrieved from the IMF:IFS database.

5. Empirical findings

The empirical findings based on both the diagnostic tests and the three hypotheses are presented in this section. Firstly, Section 5.1 presents the empirical findings considering the correctness of the dataset. Secondly, the findings for the three hypotheses are presented in Section 5.2.

5.1 Model testing

First of all, a correlation matrix is conducted to all variables to check for multicollinearity. As can be seen in Table A7, a large percentage of all variables correlate only slightly with each other. Only the geographical distance variable indicate a high correlation with the distance proxy to market centers,DIST(LF)it, of Nachum et al. (2008). Because of this near multicollinearity

the last named variable will be kept out off the dataset during the hypotheses testing, as this is the least used control variable among several studies and the least important one in this study.

Second of all, the Jarque-Bera test (Figure A4) shows that the dependent variable is not normally distributed at a 1% probability level, mostly due to an excess kurtosis of 2.01 (a normal distribution has a kurtosis of 3, which is an excess kurtosis of 0). Therefore, the dependent variable is hyperbolic secant distributed and, hence, has steeper tails. Brooks (2008: 162) mentions that this is indeed a characteristic of economic (and financial) time series. Despite of this non-normality the OLS method can still be used, as the regression asymptotically follows the appropriate distribution even in the absence of error normality (Brooks, 2008: 164).

Third of all, an overview of the descriptive statistics of all variables is included in Table A3. A remarkable observation (and outlier) is the interest rate of Brazil in January 1999 at an astonishing rate of 97.7 percent. This high percentage, together with some following high observations, was probably due to the Brazilian adoption of a floating exchange rate regime plus the currency crisis in Brazil at that moment.

Exchange Rate Movements and Foreign Direct Investments: